Re: Git help, please

[Phil Holmes]

----- Original Message ----- 
From: "Graham Percival" <address@hidden>
To: "Phil Holmes" <address@hidden>
Cc: "Devel" <address@hidden>
Sent: Wednesday, April 04, 2012 4:43 PM
Subject: Re: Git help, please


> On Wed, Apr 04, 2012 at 04:37:53PM +0100, Phil Holmes wrote:
> > Background: it seems to me the safest way to be sure that all the
> > updates I'm going to do to the snippets don't kill the documentation
> > build is to test it on a clean build (which my patchy user pretty
> > much is).  I believe the easiest way to do that will be to create a
> > patch with my normal user, upload that to a new remote branch:
> > dev/philh; then log in to patchy, pull master, pull dev/philh and do
> > a full make, make test, make doc.
>
> ...
> I'm not totally certain I follow that, but I believe you are
> incorrect.
>
> Here's what to do:
> 1) make a new branch (locally).  See CG "git for the impatient"
> for this.

I can do this :-)

> 2) run makelsr.py and point to the downloaded snippets.
>
> 3) do the security check, then commit the results.  Testing comes
> later; you can always remove the commit if it's bad.

> 4) nuke your build dir.  You've got a fast computer.
>
> 5) do a full doc build from scratch.  That'll take, what, 8
> minutes for you?  :)

About 14, including nuke, make and make doc.

> 6) if it doesn't blow up, then either push directly to staging, or
> upload to rietveld.
>
>
> Don't worry about doing extra checks with patchy.  If it passes a
> doc build from scratch on your computer, then it's highly unlikely
> to cause a problem at the patchy stage.

I have had a problem with this in the past, which is why I was suggesting 
this route.  I had an uncommitted file on my local machine that wasn't in 
master.  I'd not added it with commit -a or whatever, and so my build was OK 
but patchy-staging fell over.  I wanted to avoid the possibility.

--
Phil Holmes